Gunnar,
By far the simples way of doing this is to cheat. I normally make a separate object for each method. I can then go back and get the times from each object. Something like this:
library(xcms)
library(faahKO)
gxs<-group(faahko)
#262 325 387 450 512 575
ret<-retcor(gxs, p="d", f="s")
#Retention Time Correction Groups: 132
head(ret@peaks)
head(gxs@peaks)
Notice that the rtmed, rtmin and rtmax have shifted. This is as I say the simplest way of doing it. Otherwise you're looking at accessing the rt slot. The rt slot in xcmsSet holds both raw and corrected retention times. So xset@rt$raw has a length of the number of samples, in each of these are the original retention times from each scan. This is a bit of a pain because it means that you can't just get the peak index and call it the same as the rt$corrected/rt$raw. You need to look and see which which retention time is closer in either the raw or corrected depending on your object. I'll try a little code but it's untested so there might be error/bug/warnings.
minRT<-ret@peaks[1,"rtmin"]
maxRT<-ret@peaks[1,"rtmax"]
inx<-which(ret@rt$corrected[[ret@peaks[1,"sample]]] > minRT & ret@rt$corrected[[ret@peaks[1,"sample"]]] <maxRT )
## this is the corrected rt index so the raw should be :
ret@rt$raw[[ret@peaks[1,"sample"]]][inx]